Measurement Error Detection for Stereo Visual Odometry Integrity

Abstract

Future autonomous systems are expected to bring significant convenience to people. To ensure operational safety, navigation systems must continuously provide positioning solutions with high integrity while also meeting accuracy requirements. In addition, because those systems are applied over various scenarios, the navigation systems also must provide high robustness against environmental changes. Here, the term scenario refers to the environment around the camera (such as various light conditions, weather, and moving objects), and it does not include the subject (such as a vehicle) mounting the camera. Therefore, providing high integrity and robustness for future autonomous systems across various scenarios is one of our key research challenges. Integrity is a quantifiable performance metric used to set certifiable requirements on an individual subsystem to ensure a level of safety for the overall system (Kelly & Davis, 1994). It is a key metric